Drifting back towards Subjectivity

Definition of agenthood now depends not only on evaluation of application performance but also on protocol definition. What counts as a request?

An inexact reply is not really an answer to the question:
z.b., "Do you know the time?"
The answer is somehow "intelligent" if it is not a direct reply (exact unification with a request) and nevertheless useful, perhaps more useful, to the receiver.

So the new definition of agenthood is really a thesis about distributed "intelligence", "information", and "autonomy".
Can this be formalized?

©2006 Charles Petrie